"I see data patterns."
There's a famous scene in the movie The Sixth Sense where a young boy announces, "I see dead people." Not only was it super scary, it also didn't go over very well. The adults didn't believe him. Not that it wasn't true, they just couldn't see what he could see. Working with big data, machine learning, and predictive analytics over the last few years, I've had a similar experience when declaring, "I see data patterns!"
Really? Where?! At times like these, I'm reminded that not everyone has the ability to visualize large data sets. Which brings me to the importance of effective data visualization. For business this is not a luxury, it is critical to the success of the organization. When working with data scientists and senior engineers, it's often not necessary to transform numbers into pictures. They work with data everyday, and have a good sense of the information. They also have an intuitive understanding of probabilities and statistics that comes from years of applying it real world problems.
Across the aisle, however, sits most of the rest of the human race. And honestly, I've found many C-suite executives sit on that side of the aisle. I've worked with CEO's, COO's and CPO's that were more salesmen than scientists. So I've had to figure out how to present information so they could make intelligent decisions and take actions based on data. Yes, it's inefficient and if they were smarter it would be easier. But if you want to be a successful technical leader, you have to find a way to make information easy and accessible - to everyone. So, how do you do it? Here are four simple K's I've found that have withstood the test of time and ignorance:
1. KISS - Keep It Simple Stupid
Originally a attributed to the head of Lockheed's famous Skunkworks, KISS works everywhere. For some reason, human beings just love to complicate things! Try to avoid that inclination by remembering that in the modern world, everybody is pressed for time. We all have families and health issues, and our own work to do. Think about how you can give people the right amount of information to help them do their jobs. My general rule is that if someone doesn't understand what the visual is about in the first 15 seconds, then it's back to the drawing board.
2. Know thy data
Your selection of bar graph, line graph, or scatter plots tells a lot about your understanding of the information. But be prepared for questions about variance, confidence intervals, outliers and the like. And be open. Hard questions are the best. Don't be defensive. Let the data tell the story, and let the data be the star. I like to think of myself as a translator - my job is to try to transform data as accurately as I can and share it with as many people as possible. Before you create any visual or data representation, know in advance what you are trying to say.
"If you can't explain it simply, you don't understand it well enough." - A. Einstein
3. Keep the end in mind
There's a reason you're taking the time and energy to do this. You're trying to effectively communicate information. In order to have an action taken. One thing I try to keep in mind, is that human beings love stories. And even a data visualization works with a beginning, middle and end. Or a hypotheses, the data and a conclusion. Who's your audience? Why is this important? What actions do hope result from this information? Know in advance what conclusion you hope to draw, and what actions you hope result from the data visualization.
4. Keys and labels
Don't forget to have a key! I can't tell you how many times I've seen this rookie mistake in data visualizations and presentation. And by this I mean a key for all the data, not just the cool parts or the ones that you think are interesting. And adding labels on your data points can actually help you tell your story quickly and succinctly. Furthermore, it is much easier for your reader to read a mark label than mouse over for a tool tip.
Here's a great visual guide book from Tableau if you're interested: Simple Techniques. Good luck, may the right view be with you!
#bigdata #machinelearning #datavisualizations #datascience